kmayerb/tuna: See what may have passed through your glmnet.

Tuna makes it easy to repeat regularized regressions in the popular package [glmnet] (https://www.jstatsoft.org/article/view/v033i01) and vizualize the results. Lasso, ridge, and glmnet regressions typically use cross validation to determine a suitable value of the regularization hyperparameter that minimizes the cross-validated estimate of out of sample error. The number of non-zero coefficients can vary depending on initialization of coordinate gradient decent used to estimate arguments that maximize the penalized likelihood function. Therefore, if regularized regression is being used for feature selection it is worth repeating cross validation and gradient decent.

install.packages("remotes")
remotes::install_github("kmayerb/tuna")